We propose to develop an image-based diagnostic system for Glioblastoma (GBM) that identifies the potential genetic underpinnings of treatment resistance within the zone where tumor almost always recurs. This should facilitate the delivery of individualized care for patients with GBM. Current therapy selection is formulaic and uniform for all patients and does not account for broad genetic diversity that contributes to treatment resistance and dismal prognosis. Specifically, each patient's GBM is uniquely heterogeneous and comprised of multiple distinct subclonal populations with differing susceptibilities to therapy This diversity causes tumors to respond non-uniformly to targeted therapy and allows resistant clones to repopulate as recurrent disease. Additionally, conventional MRI routinely guides surgical resection of enhancing tumor core, but leaves behind tumor populations within adjacent non-enhancing parenchyma, or brain around tumor (BAT). The BAT represents the primary target of adjuvant therapy because it harbors the residual tumor populations that nearly universally recur. Curating the genomic diversity within BAT should inform treatment selection, but this region is almost never biopsied because it is poorly evaluated on conventional MRI. Currently, there is no systematic method that addresses intratumoral heterogeneity to characterize the regional genomic diversity within BAT. To address this critical need, this exploratory proposal will develop and test a novel mapping system that integrates multi-parametric MRI with image-guided tissue analysis and machine learning (ML) algorithms to delineate regional genomic variations in GBM. This system uses conventional MRI to identify major tumoral subcomponents: enhancing core, BAT, and central necrosis. Within these subcomponents, advanced MRI (perfusion, diffusion, texture) will further characterize regional tumor properties (i.e., angiogenesis, permeability, invasion, and proliferation) that represent phenotypic expression of underlying genomic status. These MRI traits will guide stereotactic biopsies from distinct tumoral subregions to generate matched pairs of MRI and genomic data. An ML algorithm will incorporate these data to estimate regional genomic diversity throughout each tumor, including BAT areas that have not been surgically sampled. We have unified a multi-disciplinary team of investigators from institutions that have substantial and long-standing collaborations. Our group offers expertise in multiple fields of study that are necessary to accomplish the research aims, including: 1) image processing and analytics; 2) image-guided stereotactic surgery and coregistration; 3) development of machine learning (ML) methodology; and 4) comprehensive genomic interrogation and cell biology of tumor within BAT. If successful, the work proposed here should significantly impact how GBM patients are diagnosed and treated. This potentially improves clinical outcomes by enabling a paradigm shift from one treatment fits all to a mutations-based approach that selects combinatorial therapies targeting individual tumor populations.